Mapping invasive plant species with a combination of field and remote sensing data

Invasive plant species are among the top five threats to biodiversity at a global scale. However, mapping of plants and plant communities over large areas is difficult, time consuming and expensive when using traditional field surveys. For management purposes, the early detection of invasive species is crucial. High resolution hyperspectral and LiDAR data offer a great potential to map and monitor invasive plant species, especially so at an early detection stage, and their impact on ecosystems. By combining presence-only data of the target species with airborne remote sensing data, our goals were (1) to develop an approach that allows to efficiently map biotic invasions; and (2) to create distribution maps for invasive plant species in two study areas in Western Europe, based on the aforementioned approach. This will offer the basis for further analyses of the impact of invasions on ecosystem functioning and will thus allow to infer possible management options. For this purpose we collected vegetation data on 120 plots with a size of 3 m × 3 m on the island of Sylt (Northern Germany). The plots had different cover fractions of the invasive moss Campylopus introflexus and the invasive shrub Rosa rugosa. In the forest of Compiègne (Northern France), we sampled a total of 50 plots with a size of 25 × 25 m, targeting the invasive tree Prunus serotina. In both study areas, independent validation datasets containing presence and absence points of the target species were collected. Airborne hyperspectral data were acquired for both study areas in summer 2014 by means of the APEX imaging spectrometer, providing 285 spectral bands (350-2500 nm) with a pixel size of 1.8 and 3 m, respectively. For classification, we used maxent. The results for Sylt showed that mapping invasive species using one-class classifiers is possible. For C. introflexus, the overall detection accuracy was 72%, for R. rugosa this was 92%. Most of the misclassified validation plots either had a very low cover percentage of the target species, a high cover of spectrally similar species, or were located less than one pixel away from predicted presences. In this presentation, we want to show the results of our study and evaluate the potential for mapping invasive plant species with hyperspectral and LiDAR data, especially the potential for early detection and the transferability of our models to similar study areas. Furthermore, we will discuss possible error sources connected to the collection of field and remote sensing data, as well as in the data analysis and interpretation process